Factory-Keyed Equipment: Enabling the Agile Fab

Reducing cycle time and the cost of operation is the permanent mantra of IC manufacturers across the industry. As fabs move to high product-mix manufacturing, novel approaches to automation will be critical to make fabs more agile and able to respond to the demands of current and next-generation semiconductor manufacturing.

Automation is a key aspect of achieving productivity improvements — whether those improvements are identified by the labels "agile fab" or "300 mm Prime." Improving equipment's responsiveness to manufacturing has been highlighted as one of the major focuses of increasing productivity for leading-edge manufacturing.

In the majority of today's fabs, typical fab automation architectures are implemented based on a three-tier model. At the lowest level is the equipment layer. This consists of process, metrology, an automated material handling system (AMHS) and back-end equipment. Next is the integration layer, where the bridge between the equipment and fab is provided through the implementation of station controllers. At the top is the fab application layer, where the manufacturing execution system (MES), schedule/dispatch and data analysis applications (Fig. 1 ) are typically found.

1. This three-tier fab automation model is typical in the majority of today’s fabs.

While this model has served the industry well in the past, it is not optimal moving forward. It does not provide the tight level of control required of the manufacturing process to meet expectations as to the reductions in process variability needed to meet and maintain yield and throughput requirements. The approach required now and in the future is based on the concept of factory-keyed equipment.

Factory-keyed equipment is based on the tenet that all equipment in a semiconductor fab provide the availability, access and persistence required to enable effective data management and process control. This requires a change in the way data is valued by the semiconductor industry — from simply "data" consisting of points and parameters to a representation of complex system behaviors. By understanding the resource models of the systems that comprise a semiconductor fab, data becomes the core by which decision processes are enabled (Fig. 2 ).

2. Unlike the three-tier model shown in Fig. 1 , factory-keyed equipment is two-tiered, with all equipment providing the availability and access required to enable effective data management and process control.

As has been noted over time, the critical component in fab-wide automation architectures is the station controller. It receives the lot processing information from the fab application layer and turns it into instructions that are understood by the equipment, so as to direct device fabrication. As part of its role, the station controller collects data on the fabrication process underway at the equipment and communicates this back to the fab application layer for analysis. While simple in concept, implementation is often more difficult, and the road to deployment has its share of inherent problems.

When you take a critical look at station controller implementation as part of a fab automation strategy, what you will find more often than not is that these are implemented and deployed following equipment delivery. Station controller implementation roadmaps typically follow fab ramp plans and are not often considered a forefront activity; the prevailing attitude is "we need station controllers," but that is secondary to getting equipment up, qualified and into production.

Next, we find that, for the most part, the responsibility for implementing station controllers is often assigned to a company's information technology (IT) staff or outsourced to a third-party systems integrator. While they understand computer technology and automation concepts, it is not often that they have in-depth understanding of equipment processing and how it relates to detailed production requirements. Limited experience results in an extended level of efforts to achieve basic functionality.

But the problem is deeper. Often, those responsible for building and deploying station controllers are not aware of the importance of data collected from equipment and what role it plays in assisting fab management to address critical issues, such as process variability. This expertise resides with fab equipment, process engineers and equipment manufacturers themselves. In the end, to be successful, especially as wafer sizes increase and node geometries decrease, this experience needs to be tapped upfront, prior to equipment installation, to drive fab efficiency improvements.

One area where more convergence is needed is the process of equipment qualification. Having access to real-time equipment data during the qualification process optimizes the time that it will take to release equipment to production. Using the traditional approach, station controllers are not available early enough to support this process in the majority of the cases. Moving the automation necessary for data collection to the equipment level will mitigate the current issues experienced by IC manufacturers caused by the lack of available actionable data when ramping processes (Fig. 3 ).

3. Moving the automation necessary for data collection to the equipment level will alleviate problems caused by the lack of available actionable data when ramping processes.

Another consideration that should be given when rationalizing the need for the factory-keyed equipment concept is control and exception handling. Traditional station controllers do a good job overall handling normal operating scenarios, but are not robust when it comes to exception handling at the equipment level. This results from both a lack of experience with the equipment by those implementing station controllers, and failing to develop the appropriate methods to extract this information from equipment and process engineers.

The concept of factory-keyed equipment specifically addresses these types of issues by giving the responsibility for data collection and equipment control implementation to the OEM. First, the OEM understands what data is available from the equipment and how it is to be used to ensure optimal overall equipment effectiveness (OEE) and manage the production process. Next, since they have the expertise, the OEM is best suited to work with its customers to develop the operational scenarios needed to support fab production. While the concept is clear, the path to availability needs to be planned and properly navigated for success.

When looking at making actionable data available from the equipment directly to fab-level applications, it is important to note that both legacy methods and those promoted by SEMI e-manufacturing standards need to be supported. This data needs to be made available on demand to those applications that will analyze the data and then determine what needs to be done at the equipment level to meet target production goals. By removing the role of data collection from the station controller and driving this to the equipment level, the needed results may be achieved.

Some will argue that this is difficult because there may be many different data collection requirements, depending on the fab-level applications deployed. Further, it may be argued that to support the different types of fab-level applications would require a significant level of effort. While it may appear this way on the surface, the reality is that the ability to extract and distribute data to different applications is not difficult. Using modern software technology and methods, the adoption of SEMI e-manufacturing standards drives this capability to be a configurable process, and is thus easily adapted by OEMs to meet their different customer requirements.

So, the next question is who is responsible for implementation? The ideal situation is one where the specific configuration of data collected from the equipment is defined by the IC manufacturer's process and equipment engineering staff. They will be able to directly set up what data they want access to and what actions should be taken on the collected data. This approach empowers the domain knowledge of the IC manufacturer and shortens the time required to get the data needed to support manufacturing. In addition, providing the IC manufacturer with the ability to manage data collection directly from the equipment gives manufacturers the flexibility they need to make changes rapidly based on observed conditions.

Another consideration is how to address operational scenarios. These scenarios have, in the past, been implemented in station controllers as custom logic. They are often "hard-coded" and specific to the IC manufacturer and equipment being integrated. In the past, these station controllers have become what are known as "thick clients" — monolithic solutions tightly coupled to the processing environment and not agile in the support of changing requirements and conditions. The factory-keyed equipment approach is based on the implementation of what are known as "thin clients," which implement processing agents for common tasks and provide the ability to configure operational scenarios to meet specific IC manufacturer requirements.

Driving the operational scenarios down to the equipment is new to the OEM, and initial resistance to this approach is expected. From a development perspective, it will require the OEM to spend the necessary time to acquire and/or implement the appropriate applications to support providing a host interface for control. It will be the position of many that this should be left to the IC manufacturer as it is today, but moving this logic to the equipment and implementing those agents specific to the equipment by the OEM will enable the IC manufacturer to achieve quick implementation and deployment.

An initial implementation of the factory-keyed equipment concept is offered as an option for the Spartan Sorter. As shown in Figure 4 , the factory-keyed equipment configuration adds two resident applications: a factory data integration manager (FDIM) and a thin-client station controller. The FDIM provides data on demand to any fab-level application and supports legacy communication via SECS/GEM, as well as support for the SEMI e-manufacturing standards known as Interface A. The thin-client station controller provides a job-step library that is fully configurable to implement operational scenarios for both normal processing and exception handling using the concepts of process job/control job, as defined by SEMI standards.

4. The factory-keyed equipment configuration adds two resident applications: a factory data integration manager (FDIM) and a thin-client station controller.

FDIM supports two modes for enabling data collection. A process or equipment engineer can configure data collection plans using an offline utility, or a fab-level application can define a data collection plan in requests sent to the FDIM. In either case, the protocol used for communication is hidden and does not require any level of effort from the IC manufacturer to configure. While FDIM supports most protocols in use today, custom interfaces can be developed as required using configuration utilities.

The addition of a thin-client station controller provides a rich library of generic job steps. Each job step has a specific action it takes, such as requesting a lot move in at the MES or recipe selection on the equipment, among others. The library of job steps has been organized based on typical manufacturing flows and provides for lot introduction, set up, processing and completion specific to the equipment and addresses both normal and exception scenarios.

Customers who implement factory-keyed equipment may configure their operational scenarios using available utilities or work with application specialists to build customer scenarios in advance so that they are available out of the box when the equipment arrives. In either case, getting fully functioning equipment running in a production environment may be completed as much as six to eight weeks earlier than using traditional station controller approaches.

The question that remains is if factory-keyed equipment is a benefit to IC manufacturers and OEMs, then why hasn't it been widely embraced by the semiconductor industry? The answer is that barriers to entry are high in semiconductor manufacturing, and market pressures drive the adoption of displacing technologies. As more IC manufacturers strive to make their fabs more agile and request the benefits available through factory-keyed equipment, then more solutions will be developed for the industry. As is the case in general, unless the customer is requesting it, there is no incentive to build it.

Over time and with the appropriate forums established within venues such as SEMI and Sematech, a migration to factory-keyed equipment automation architectures will gain momentum. But as with all automation advances within the semiconductor industry, this will require more discussion and consensus between IC manufacturers and OEMs. As the trend toward high product-mix facilities continues, the question is not "if" factory-keyed equipment will become a de-facto approach in the future, it "when." Those that lead and support this initiative will be the first to gain from its adoption.